基于通用学习均衡优化器的多阈值图像分割  

Multi Threshold Image Segmentation Based on General Learning Equalization Optimizer

作　　者：吴佳芸 武灵芝[1] 胡晓飞[1] WU Jiayun;WU Lingzhi;HU Xiaofei(School of Geography and Biological Information,Nanjing University of Posts and Telecommunications,Nanjing Jiangsu 210023,China)

机构地区：[1]南京邮电大学地理与生物信息学院,江苏南京210023

出　　处：《传感技术学报》2024年第3期463-468,共6页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(61771251);江苏省自然科学基金项目(BK20171443)。

摘　　要：传统的元启发式多阈值图像分割算法计算复杂度高且容易陷入局部最优,通用学习均衡优化器在搜索过程中使粒子从不同维度的候选粒子中学习,在求解复杂问题最优解时有很强的能力,克服了容易陷入局部最优的问题。提出将通用学习均衡优化算法优化最大类间方差法来实现多阈值图像分割,实验选择标准灰度图像,以峰值信噪比、结构相似度、运行时间和适应度值为评价标准,将该算法与均衡优化算法、粒子群优化算法进行了比较。结果表明,基于通用学习均衡优化器的多阈值图像分割算法结果的峰值信噪比、结构相似度在绝大多数情况下优于另外两个算法,并且收敛速度快,执行效率高。The traditional meta-heuristic multi-threshold image segmentation algorithm has high computational complexity and is easy to fall into local optimum,whereas the general learning equilibrium optimizer enables particles to learn from candidate particles in different dimensions during the search process having a strong ability in solving the optimal solution of complex problems and overcoming the problem of easily falling into local optimum.The general learning equalization optimization algorithm is proposed to optimize the maximum inter-class variance method to realize multi-threshold image segmentation,and the standard grayscale images are selected in the experiment.Taking the peak signal-to-noise ratio,structural similarity,running time and fitness value as the evaluation criteria,the algorithm is compared with the equalization optimization algorithm and the particle swarm optimization algorithm.The results show that the peak signal-to-noise ratio and structural similarity of the multi-threshold image segmentation algorithm based on the general learning equalization optimizer are better than the other two algorithms in most cases,the convergence speed is fast and the execution efficiency is high.

关 键 词：数字图像处理 多阈值图像分割 通用均衡优化器 最大类间方差法 粒子群优化算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]